[PATCH i-g-t v4 1/2] lib/igt_kmemleak: library to interact with kmemleak

[Peter Senna Tschudin]

On 12.02.2025 14:41, Kamil Konieczny wrote:
> Hi Peter,
> On 2025-02-12 at 14:12:44 +0100, Peter Senna Tschudin wrote:
>>
>>
>> On 12.02.2025 11:01, Zbigniew Kempczyński wrote:
>>> On Wed, Feb 12, 2025 at 10:18:17AM +0100, Peter Senna Tschudin wrote:
> 
> [...]
> 
>>>> Resume: I prefer to wait until the initial feedback from CI and developers.
>>>> I will be happy to move this to runner later if we decide to not walk the
>>>> unknown road of trying to catch as many transient leaks as we can.
>>>>
>>>> That being said, let me know how to move forward.
>>>
>>> Question is to CI folks, for me this scan should be executed as
>>> dedicated CI run with kmemleak on. And I repeat - I don't see reason
>>> we should add this support to the tests, as few line script which
>>> wraps around test by kmemleak clear/scan is enough. We already have
>>> scripts directory so it is good place for such wrapper.
>>
>> So I am moving the code to the runner directory, and using printf() instead
>> of igt_warn(). Any other request for the series?
>>
>> [...]
>>
> 
> or fprintf() or write() so you could write to a file instead
> of stdout, as stdout isn't collected now by runner itself (it
> goes to igt_runnerNN.txt but that is another story).

The errors in question here are exactly about writing to disk, so
I am positive that it only makes sense for them to go to stdout or
stderr. Are you ok with printing an error message instead of trying
to save to disk?